Thomas, I think it's better to talk over call because cluster solution is really a BIG one. Some time next week? Feel free to fix a call here: https://www.gluu.org/booking

Note: Community support is limited to forums.

>>_Is there a specific load balancer that Gluu recommends for use _ >>_ with Gluu CE HA cluster? HAProxy? LVS? Linux HA? Red Hat HA?_ No, you can use any LB. >> _Do the load balancer nodes have to be separate from the _ >>_ Gluuservers or can they be installed on the same Linux server _ >>_ instances as the Gluu nodes?_ No opinion, although if you host on the same server as the Gluu Server, make sure you use a different ethernet interface of course. >>_ Is there any more specific documentation or configuration _ >>_ examples for the load balancer configuration?_ No, LB is outside of scope for Gluu. You can use any load balancer algorithm for oxAuth, as it uses LDAP to distribute state. oxTrust has web sessions. If you load balance it, you'll need to use sticky sessions. Or even better, don't make it Internet facing. >>_ What ports are required to be handled by the load balancers? Just :443? _ Yes. >> _Can an existing, working single-node Gluu CE deployment be migrated to an HA cluster by adding an additional node?_ As long as you setup multi-master replication, and the IP networking works (i.e. server 1 can communicate with server 2) >> _How does running a Gluu CE HA cluster affect the upgrade process?_ The process applies: 1. Export data 2. Create new server with old `setup.properties.last` 3. Import data 4. Test 5. Do cutover 6. Test 7. Revert or turn off old instance.

Thanks for the info. I don't think we need to do a phone call. Also, for future references/searches, there's some good info on Shib idP clustered environments here: [https://wiki.shibboleth.net/confluence/display/SHIB2/IdPClusterIntro](https://wiki.shibboleth.net/confluence/display/SHIB2/IdPClusterIntro)

Cool! Post here if you are successful.

Hi, I backburner-ed this project for a long while, but I am working on it now with the improved documentation in 3.1.2 for community clustering. I do have a question. In the 3.1.2 clustering documentation, the example states that the writer ran his redis instance on one of the redundant gluu instances, outside the chroot container. My question is what will happen if that gluu host goes down? How does redis disappearing affect the application? My assumption is it won't function without redis, and so any gluu host (outside chroot) hosting redis will be unable to go down. If this is true, I will just run redis on a separate server. I looked into running redis with sentinal cluster but the number of hosts it requires is too much complexity. I think we can live with a single point of failure in redis as long as it it's own separate host.